samba - Oct 2017 - transfer hanging to older smb device

Hi,

I know this is not a samba _server_ issue but it is still an issue with 
my cifs.ko apparently...

I have mounted over smb/cifs an older Freecom network drive.

The mount is done on an older device as well, a Synology DiskStation 
running kernel 2.6.32 and using mount.cifs version 6.5 that I compiled 
myself.

I am writing files to this mount, and while writing (using dd) the copy 
will just stall and stop apparently doing anything.

If I just interrupt the process and restart after a while, nothing 
happens.

"vmstat" is not available on this device.

If I "ls" the mount then it takes a few seconds to respond, and then 
immediately continues writing.

Now it complains that the host is down, but my other computer can access 
the share just fine...

As well as that I can ping the machine from the Synology.

Maybe it is getting caused by DD using a too high blocksize...

Now it has restarted again....

Xen via samba schreef op 05-10-2017 5:14:
> Maybe it is getting caused by DD using a too high blocksize...
I think it was caused by using a DD bs of 4M.

I put it back to 128k now and there doesn't seem to be an issue anymore.

The Linux Kernel is generally terrible in the area of IO queues.

---------------------------------

I was meaning to ask something else though.

When my DD copy to the share completes, it often ends with an error:

"closing output file 'XXX': Input/output error"

 From a bug on NFS[1] it appears to be related to short I/O timeouts.

Is there anything I can do to mitigate that?


I don't want DD to close with an error.

I am using completely general mount options.

There is apparently sometimes an IO timeout when it closes the file.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=668945